🔍 Cross-Domain Sentiment Analysis

This repository demonstrates a comprehensive cross-domain sentiment analysis project covering multiple datasets, preprocessing pipelines, model development, cross-domain evaluation, and visualization of results.

The goal of this project is to evaluate how well sentiment classification models trained on one domain generalize to other domains.

---

📚 Project Overview

This project works with sentiment datasets from different domains, including:

• Movie reviews (IMDB)
• Social media posts
• Twitter airline sentiment
• Twitter training dataset
• Consumer complaint narratives

We explore:

• Dataset-specific preprocessing strategies
• TF-IDF–based feature extraction
• Logistic Regression–based classifiers
• Cross-domain model transferability
• Performance comparison through visualizations

---

📁 Repository Structure

.
├── Preprocessing/
│   ├── imdb_preprocessing.ipynb
│   ├── socialmedia_preprocessing.ipynb
│   ├── twitter_us_airline_preprocessing.ipynb
│   ├── twitter_training_preprocessing.ipynb
│   └── consumer_complaints_preprocessing.ipynb
│
├── clean_data/
│   ├── IMDB-Dataset_cleaned.csv
│   ├── socialmedia_sentiment_cleaned.csv
│   ├── twitter_us_airline_cleaned.csv
│   ├── twitter_training_cleaned.csv
│   └── consumer_complaints_cleaned.csv
│
├── plots/
│   └── (Saved plots such as confusion matrices, accuracy comparisons)
│
├── model_development.ipynb
├── cross_domain_analysis_with_visualization.ipynb
└── README.md

---

Workflow Description

🔹 Data Preprocessing

Each dataset is cleaned and preprocessed independently to handle:

• Text normalization
• Noise removal
• Label consistency
• Domain-specific characteristics

All preprocessing notebooks are located in the Preprocessing/ folder, and the cleaned datasets are saved to clean_data/.

---

🔹 Model Development

The model_development.ipynb notebook performs:

• TF-IDF feature extraction
• Model training using Logistic Regression
• Saving trained models and vectorizers for reuse

Each dataset is treated as a separate training domain.

---

🔹 Cross-Domain Evaluation

The cross_domain_analysis_with_visualization.ipynb notebook:

• Loads trained models and vectorizers
• Evaluates them on datasets from other domains
• Computes accuracy, confusion matrices, and comparison metrics
• Generates visualizations for performance analysis

All plots are saved in the plots/ directory.

---

▶️ How to Run

1. Clone the repository

   git clone https://github.com/<your-username>/Cross-Domain-Sentiment-Analysis.git
   cd Cross-Domain-Sentiment-Analysis

2. Preprocess datasets

   • Open notebooks inside the Preprocessing/ folder
   • Run each notebook to generate cleaned datasets

3. Train models

   • Open and run model_development.ipynb

4. Run cross-domain analysis

   • Open and run cross_domain_analysis_with_visualization.ipynb
   • Generated plots will be saved automatically in plots/

---

📊 Outputs

• Cleaned datasets for each domain
• Trained sentiment classification models
• Cross-domain performance metrics
• Visualizations including:
• Confusion matrices
• Accuracy comparison plots

---

🔧 Requirements

• Python 3.x
• Pandas
• NumPy
• Scikit-learn
• Matplotlib
• Seaborn
• Jupyter Notebook or Google Colab

---

👥 Team Members

• Arpitha Thippeswamy
• Anti Li

---